make trait-aliases work across crates
This is rebase of a small part of @alexreg's PR #55994. It focuses just on the changes that integrate trait aliases properly into crate metadata, excluding the stylistic edits and the trait objects.
The stylistic edits I also rebased and can open a separate PR.
The trait object stuff I found challenging and decided it basically needed to be reimplemented. For now I've excluded it.
Since this is really @alexreg's work (I really just rebased) I am going to make it r=me once it is working.
Fixes#56488.
Fixes#57023.
Fix stack overflow when finding blanket impls
Currently, SelectionContext tries to prevent stack overflow by keeping
track of the current recursion depth. However, this depth tracking is
only used when performing normal section (which includes confirmation).
No such tracking is performed for evaluate_obligation_recursively, which
can allow a stack overflow to occur.
To fix this, this commit tracks the current predicate evaluation depth.
This is done separately from the existing obligation depth tracking:
an obligation overflow can occur across multiple calls to 'select' (e.g.
when fulfilling a trait), while a predicate evaluation overflow can only
happen as a result of a deep recursive call stack.
Fixes#56701
I've re-used `tcx.sess.recursion_limit` when checking for predication evaluation overflows. This is such a weird corner case that I don't believe it's necessary to have a separate setting controlling the maximum depth.
Currently, SelectionContext tries to prevent stack overflow by keeping
track of the current recursion depth. However, this depth tracking is
only used when performing normal section (which includes confirmation).
No such tracking is performed for evaluate_obligation_recursively, which
can allow a stack overflow to occur.
To fix this, this commit tracks the current predicate evaluation depth.
This is done separately from the existing obligation depth tracking:
an obligation overflow can occur across multiple calls to 'select' (e.g.
when fulfilling a trait), while a predicate evaluation overflow can only
happen as a result of a deep recursive call stack.
Fixes#56701
When we coerce `dyn Foo` to `dyn Bar`, that is OK as long as `Foo` is
usable in all contexts where `Bar` is usable (hence using the source
must be a subtype of the target).
This is needed for the universe-based code to handle
`old-lub-glb-object`; that test used to work sort of by accident
before with the old code.
Previously, evaluation ignored outlives relationships. Since we using
evaluation to skip the "normal" trait selection (which enforces
outlives relationships) this led to incorrect results in some cases.
The future-compat warnings break code that assumes that `dyn Send + Sync !=
dyn Sync + Send`, and are the first step in making them equal. cc #33140.
It should be possible to revert this commit when we're done with the
warnings.
Take 2: Implement object-safety and dynamic dispatch for arbitrary_self_types
This replaces #50173. Over the months that that PR was open, we made a lot of changes to the way this was going to be implemented, and the long, meandering comment thread and commit history would have been confusing to people reading it in the future. So I decided to package everything up with new, straighforward commits and open a new PR.
Here are the main points. Please read the commit messages for details.
- To simplify codegen, we only support receivers that have the ABI of a pointer. That means they are builtin pointer types, or newtypes thereof.
- We introduce a new trait: `DispatchFromDyn<T>`, similar to `CoerceUnsized<T>`. `DispatchFromDyn` has extra requirements that `CoerceUnsized` does not: when you implement `DispatchFromDyn` for a struct, there cannot be any extra fields besides the field being coerced and `PhantomData` fields. This ensures that the struct's ABI is the same as a pointer.
- For a method's receiver (e.g. `self: Rc<Self>`) to be object-safe, it needs to have the following property:
- let `DynReceiver` be the receiver when `Self = dyn Trait`
- let `ConcreteReceiver` be the receiver when `Self = T`, where `T` is some unknown `Sized` type that implements `Trait`, and is the erased type of the trait object.
- `ConcreteReceiver` must implement `DispatchFromDyn<DynReceiver>`
In the case of `Rc<Self>`, this requires `Rc<T>: DispatchFromDyn<Rc<dyn Trait>>`
These rules are explained more thoroughly in the doc comment on `receiver_is_dispatchable` in object_safety.rs.
r? @nikomatsakis and @eddyb
cc @arielb1 @cramertj @withoutboats
Special thanks to @nikomatsakis for getting me un-stuck when implementing the object-safety checks, and @eddyb for helping with the codegen parts.
EDIT 2018-11-01: updated because CoerceSized has been replaced with DispatchFromDyn
I don't really understand what it's for, but see the comment here:
https://github.com/rust-lang/rust/pull/50173#discussion_r204222336
where arielb1 said
> Does this check do anything these days? I think `$0: Trait` is always considered ambiguous
and nikomatsakis agreed we may be able to get rid of it
Fix Rustdoc ICE when checking blanket impls
Fixes#55001, #54744
Previously, SelectionContext would unconditionally cache the selection
result for an obligation. This worked fine for most users of
SelectionContext, but it caused an issue when used by Rustdoc's blanket
impl finder.
The issue occured when SelectionContext chose a ParamCandidate which
contained inference variables. Since inference variables can change
between calls to select(), it's not safe to cache the selection result -
the chosen candidate might not be applicable for future results, leading
to an ICE when we try to run confirmation.
This commit prevents SelectionContext from caching any ParamCandidate
that contains inference variables. This should always be completely
safe, as trait selection should never depend on a particular result
being cached.
I've also added some extra debug!() statements, which I found helpful in
tracking down this bug.
Fixes#55001, #54744
Previously, SelectionContext would unconditionally cache the selection
result for an obligation. This worked fine for most users of
SelectionContext, but it caused an issue when used by Rustdoc's blanket
impl finder.
The issue occured when SelectionContext chose a ParamCandidate which
contained inference variables. Since inference variables can change
between calls to select(), it's not safe to cache the selection result -
the chosen candidate might not be applicable for future results, leading
to an ICE when we try to run confirmation.
This commit prevents SelectionContext from caching any ParamCandidate
that contains inference variables. This should always be completely
safe, as trait selection should never depend on a particular result
being cached.
I've also added some extra debug!() statements, which I found helpful in
tracking down this bug.
